类似如下：

[23.11|08:46:11]     4       45 FAILED   TEMPERATURE                                        
[23.11|08:46:11]     4       46 FAILED   TEMPERATURE                                        
[23.11|08:46:11]     4       47 FAILED   TEMPERATURE                                        
[23.11|08:46:11] --- End summary ---
[23.11|08:46:11] 
[23.11|08:46:11] Running more reference calculations.... 
[23.11|08:46:11]     For job step4_attempt47_simulation, temperature > 5000.0 K starting at frame 37
[23.11|08:46:11]     Running reference calculations on frames [36] from step4_attempt47_simulation\ams.rkf
[23.11|08:46:11]     Calculating 1 frames in total
[23.11|08:46:11]     Running step4_attempt47_reference_calc2
[23.11|08:48:41]     Reference calculations finished!
[23.11|08:48:41] WARNING: Reference job step4_attempt47_reference_calc2 failed: UNKNOWN. Attempting to use data anyway...
[23.11|08:48:41] Traceback (most recent call last):
  File "C:\Users\testadf\AppData\Local\Temp\pip-target-8dm09ykr\lib\python\scm/simple_active_learning/simple_active_learning_workflow.py", line 882, in run
  File "C:\Users\testadf\AppData\Local\Temp\pip-target-8dm09ykr\lib\python\scm/simple_active_learning/simple_active_learning_workflow.py", line 714, in run
  File "C:\Users\testadf\AppData\Local\Temp\pip-target-8dm09ykr\lib\python\scm/simple_active_learning/active_learning.py", line 355, in run_loop
  File "C:\Users\testadf\AppData\Local\Temp\pip-target-8dm09ykr\lib\python\scm/simple_active_learning/active_learning.py", line 449, in update_and_store_singlepoint_reference_data
  File "C:\Users\testadf\AppData\Local\Temp\pip-target-8dm09ykr\lib\python\scm/simple_active_learning/utils.py", line 101, in add_singlejob
  File "C:\Users\testadf\AppData\Local\Temp\pip-target-8dm09ykr\lib\python\scm/params/core/results_importer.py", line 345, in add_singlejob
TypeError: unsupported operand type(s) for *: 'NoneType' and 'float'

[23.11|08:48:41] ERROR: unsupported operand type(s) for *: 'NoneType' and 'float'
Job TIH2+O2 has finished

从logfile可以看到训练到某一个分段（一般默认10个分段，在logfile中叫做Step），尝试了47次训练（在logfile中叫做attempt，因此合起来叫做step4_attempt47，即第四段，训练了47次都训练不出来），都没有训练好。

用最后一次训练的结果，即logfile最后一次出现的训练结果，例如：

Engine MLPotential
  Backend M3GNet
  MLDistanceUnit angstrom
  MLEnergyUnit eV
  Model Custom
  ParameterDir E:\WJC-ML\TIH2+O2\TIH2+O2.results\step4_attempt46_training\results\optimization\m3gnet\m3gnet
EndEngine

作为初始，继续往下重新训练即可。

训练的结果，如上例子，则保存在E:\WJC-ML\TIH2+O2\TIH2+O2.results\step4_attempt46_training\results\optimization\m3gnet\m3gnet（用户根据自己的实际位置去设置，这里只是一个案例而已）。关于如何基于一个已有结果继续训练，请参考教程第三节：AMS软件：从AIMD从头训练或微调机器学习势M3GNet、NequIP